Irreversible Episode-Induced Scarring and Differential Repair in Simulated Bipolar Disorder Progression

Background: Bipolar depression treatment is complicated by risks of manic switch and potential illness progression via kindling-like sensitization. Emerging rapid-acting agents (ketamine, neurosteroids) differ mechanistically from traditional monoaminergic antidepressants, but their long-term effects on vulnerability remain unclear. We developed a computational neural network model to compare acute efficacy, manic conversion risk, post-discontinuation relapse, and multi-cycle kindling across three mechanisms. Methods: Feedforward networks were trained on a four-class blob classification task, aggressively pruned (95% sparsity), and subjected to uniform early adversity scarring (mean 3%). Independent copies received ketamine-like (moderate gain + gradient-guided regrowth), SSRI-like (progressive high gain, no repair), or neurosteroid-like (low gain + strong inhibition) interventions. Depressive impairment was modeled via internal noise; manic conversion via biased excitatory noise. Longitudinal relapse and kindling (six cycles with weakening triggers and permanent scarring upon relapse) were simulated across 10 seeds. Results: All mechanisms restored acute performance under stress (neurosteroid-like 97.6%, ketamine-like 97.1%, SSRI-like 90.3%), but SSRI-like showed highest manic conversion risk and near-universal post-discontinuation relapse (98.3%). Kindling revealed stark divergence: SSRI-like networks sustained high relapses (3.9 total) with 30% autonomy; neurosteroid-like limited scarring (final 3.7%) but required ongoing administration; ketamine-like tolerated highest scarring (7.3%) yet achieved fewest relapses (0.7) and no autonomy via compensatory regrowth. Conclusions: Plasticity-enhancing mechanisms uniquely resist sensitization and autonomy despite cumulative damage, suggesting potential disease-modifying effects. Monoaminergic excitation may exacerbate progression in vulnerable systems. These findings highlight repair capacity as a critical determinant of long-term outcome and support prioritizing rapid-acting agents in high-risk bipolar depression.